navigator.userAgent may return outdated value after webView.customUserAgent is set, due to its cached not getting invalidated.
<rdar://problem/42566456>
Created attachment 345768 [details] Patch
Comment on attachment 345768 [details] Patch Prior to <https://trac.webkit.org/changeset/232159/webkit>, was the UI client able to change navigator.userAgent while the page was live? I'm not sure it's important or desirable to be able to change navigator.userAgent while the page is live. It's specified as a readonly attribute, and a website might do something really weird if feature detection returned different results half-way through webpage rendering.
(In reply to Geoffrey Garen from comment #3) > Comment on attachment 345768 [details] > Patch > > Prior to <https://trac.webkit.org/changeset/232159/webkit>, was the UI > client able to change navigator.userAgent while the page was live? Yes, since there was no caching whatsoever. > I'm not > sure it's important or desirable to be able to change navigator.userAgent > while the page is live. It's specified as a readonly attribute, and a > website might do something really weird if feature detection returned > different results half-way through webpage rendering. I am not sure how important this is either but this is a recent behavior change compared to shipping and I have a radar about it. Clearly, the developer who filed the radar see this as a regression.
(In reply to Chris Dumez from comment #4) > (In reply to Geoffrey Garen from comment #3) > > Comment on attachment 345768 [details] > > Patch > > > > Prior to <https://trac.webkit.org/changeset/232159/webkit>, was the UI > > client able to change navigator.userAgent while the page was live? > > Yes, since there was no caching whatsoever. > > > I'm not > > sure it's important or desirable to be able to change navigator.userAgent > > while the page is live. It's specified as a readonly attribute, and a > > website might do something really weird if feature detection returned > > different results half-way through webpage rendering. > > I am not sure how important this is either but this is a recent behavior > change compared to shipping and I have a radar about it. Clearly, the > developer who filed the radar see this as a regression. My first instinct was also not to fix it since I am not convinced there is a great use-case for changing the user agent after the page has loaded. That said, the radar was sent to me so I fixed it. It is not a lot of code and I have already written it, I do not see much harm in landing it.
Created attachment 345772 [details] Patch
(In reply to Chris Dumez from comment #5) > (In reply to Chris Dumez from comment #4) > > (In reply to Geoffrey Garen from comment #3) > > > Comment on attachment 345768 [details] > > > Patch > > > > > > Prior to <https://trac.webkit.org/changeset/232159/webkit>, was the UI > > > client able to change navigator.userAgent while the page was live? > > > > Yes, since there was no caching whatsoever. > > > > > I'm not > > > sure it's important or desirable to be able to change navigator.userAgent > > > while the page is live. It's specified as a readonly attribute, and a > > > website might do something really weird if feature detection returned > > > different results half-way through webpage rendering. > > > > I am not sure how important this is either but this is a recent behavior > > change compared to shipping and I have a radar about it. Clearly, the > > developer who filed the radar see this as a regression. > > My first instinct was also not to fix it since I am not convinced there is a > great use-case for changing the user agent after the page has loaded. That > said, the radar was sent to me so I fixed it. It is not a lot of code and I > have already written it, I do not see much harm in landing it. Note that would also impact changing the user-agent via the Develop menu (navigator.userAgent would not change until reloading).
Comment on attachment 345772 [details] Patch Clearing flags on attachment: 345772 Committed r234211: <https://trac.webkit.org/changeset/234211>
All reviewed patches have been landed. Closing bug.